City, University of London Institutional Repository Citation: Triep, M., Brücker, C., Kerkhoffs, W., Schumacher, O. & Marseille, O. (2008). Investigation of the washout effect in a magnetically driven axial blood pump. Artificial Organs, 32(10), pp. 778-784. doi: 10.1111/j.1525-1594.2008.00630.x This is the accepted version of the paper. This version of the publication may differ from the final published version. Permanent repository link: http://openaccess.city.ac.uk/12960/ Link to published version: http://dx.doi.org/10.1111/j.1525-1594.2008.00630.x Copyright and reuse: City Research Online aims to make research outputs of City, University of London available to a wider audience. Copyright and Moral Rights remain with the author(s) and/or copyright holders. URLs from City Research Online may be freely distributed and linked to. City Research Online: http://openaccess.city.ac.uk/ [email protected]City Research Online
Transcript
City, University of London Institutional Repository
Citation: Triep, M., Brücker, C., Kerkhoffs, W., Schumacher, O. & Marseille, O. (2008). Investigation of the washout effect in a magnetically driven axial blood pump. Artificial Organs, 32(10), pp. 778-784. doi: 10.1111/j.1525-1594.2008.00630.x
This is the accepted version of the paper.
This version of the publication may differ from the final published version.
Link to published version: http://dx.doi.org/10.1111/j.1525-1594.2008.00630.x
Copyright and reuse: City Research Online aims to make research outputs of City, University of London available to a wider audience. Copyright and Moral Rights remain with the author(s) and/or copyright holders. URLs from City Research Online may be freely distributed and linked to.
City Research Online: http://openaccess.city.ac.uk/ [email protected]
*Institut für Mechanik und Fluiddynamik, TU Bergakademie Freiberg, and †CiruLite Inc., Aachen, Germany
Abstract: For a long-term implementation of the magneti-cally driven CircuLite blood pump system, it is extremelyimportant to be able to ensure a minimum washout flow inorder to avoid dangerous stagnation regions in the gapbetween the impeller and the motor casing as well as nearthe pivot–axle area at the holes in the impeller’s hub. Ingeneral, stagnation zones are prone to thrombus formation.Here, the optimal impeller/motor gap width will be deter-mined and the washout flow for different working condi-
tions will be quantitatively calculated. The driving force forthis secondary flow is mainly the strong pressure differencebetween both ends of the gap. Computational fluid dynam-ics (CFD) and digital particle image velocimetry (DPIV)will be used for this analysis. Key Words: Axial bloodpump—Sealless—Magnet drive—Computational fluiddynamics—Digital particle image velocimetry—Washoutflow.
The design of miniature rotary blood pumps isoften based on established pump concepts from clas-sical turbomachinery. However, key aspects like thedifferent medium and flow regime have to be con-sidered in the early design stage for the successespecially of blood pump systems for long-termapplications. Classical impellers are driven through asealed shaft by an electric motor. The interfacebetween moving and fixed parts is always a criticalregion in rotary blood pumps, because shear strainbecomes locally very high. This leads to a higherblood traumatization close to the shaft in the axialimpeller–motor gap, and potentially to thrombus for-mation, which should be prevented in any operatingcase of the pump. This problem can be circumventedby the use of a fixed axle with an axial pivot bearingon top of it inside the impeller body and the impellerbeing driven magnetically (1,2). In combination with
two washout holes in the impeller’s hub, thrombusformation can thus be avoided at the nonrotating/rotating interface (3). This concept is used byCircuLite (Saddle Brook, NJ, USA) for an implant-able long-term microaxial blood pump that connectsthe left atrium with the arteria subclavia (Fig. 1).
The main goal of the present investigation is thestudy and optimization of the induced washout flowthrough the axial impeller–rotor gap and the radialimpeller–axle gap in order to avoid stagnation zonesnear the axle pivot during all operation conditions ofthe CircuLite pump.
Both numerical (computational fluid dynamics[CFD]) and experimental (digital particle imagevelocimetry [DPIV]) methods are used comple-mentarily. Washout-flow efficiency in the impellercould be improved through variation of the axial gapwidths.
MATERIALS AND METHODS
The MicroVad pumpThe two-bladed CircuLite impeller was incorpo-
rated into the MicroVad pump system. Blood wassucked through an inflow cannula by the impeller,which has a blade tip diameter of 6.6 mm. The flowwas redirected by the blades into the annular channel
doi:10.1111/j.1525-1594.2008.00630.x
Received February 2008; revised June 2008.Address correspondence and reprint requests to Dr. Michael
Triep, TU Bergakademie Freiberg—Institut für Mechanik undFluiddynamik, Lampadiusstr. 4 Freiberg 09596, Germany. E-mail:[email protected]
Presented in part at the 15th Congress of the InternationalSociety for Rotary Blood Pumps held on November 2–4, 2007 inSydney, Australia.
JOBNAME: No Job Name PAGE: 2 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
between the electric motor and the casing, and wasdelivered tangentially into the outlet graft. Thesystem reached at physiological conditions a dis-charge rate of 3.5 L/min and may be operated at amaximum rotational speed of 28.000 rpm. The powerconsumption was in the order of 6–9 W.
The impeller was driven magnetically so that nosealing is needed. It was mounted on top of a fixedcylindrical axle, thereby forming a constant radial gapas shown in Fig. 2. In combination with the rotor/motor gap and two holes in the impeller’s hub, awashout channel was created. During operation of
the pump, a bypass flow through this washoutchannel was induced so that thrombus formation atthe critical ball pivot is prevented.
Experimental flow circuitFor the purpose of the study of the overall perfor-
mance of the pump, in particular the visualization ofthe washout flow, the impeller was integrated into ahydraulic mock loop (Fig. 3).
A major feature of the setup is the fully transpar-ent Perspex housing of the pump, which allows fulloptical accessibility into the inlet cannula, the impel-ler, and the motor region of the main flow. The testchamber is an identical copy of the real clinical pumpdevice. A Newtonian test fluid composed of water/glycerine mixture (35 weight percent glycerine) witha density of 1060 kg/m3 and a dynamic viscosity of3.6 cP was used as working fluid. The pressure loadand flow was regulated by a throttle valve down-stream of the pump, and the flow rate was measuredby an ultrasound flowmeter. The connector of theimpeller to the mock circuit allows an easy exchangeof similar impeller types with small geometric param-eter changes.
DPIV setupDetailed measurements of the internal flow within
the pump were carried out using DPIV. This methodallows in principle to determine the planar flow fieldin a selected plane of the flow. Therefore, the laserbeam of a phase-triggered double-pulse Nd:Yag laser(Photonics, Edinburgh, UK) with a power of 30 mJ/pulse and a wavelength of 532 nm was expanded andfocused to a 0.2-mm light sheet. A prism system re-directed the sheet from top and bottom into the testchamber. A micropositioning stage allows to move
FIG. 1. Implementation of the MicroVad pump system into theleft atrium.
Ball Pivo
Washout Channel Impeller / Motor Gap
FIG. 2. Induction of the washout flow.
FIG. 3. Flow circuit for pump tests.
M. TRIEP ET AL.2
Artif Organs, Vol. ••, No. ••, 2008
4 4
12
3
456789
101112131415
16
17
18
19
202122
23
2425262728293031323334353637383940414243
44
4546474849505152535455
m101
colour online & in print
m101
colour online & in print
m101
colour online & in print
JOBNAME: No Job Name PAGE: 3 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
the whole test chamber through the light sheet. Thefull optical arrangement can be seen in Fig. 4.
The flow was seeded with small fluorescent tracerparticles (Dantec Dynamics A/S, Skovlunde,Denmark) with a mean diameter of 5 m in a concen-tration of 0.6 g/L. The fluorescent light of the par-ticles passed an orange filter (LOT-Oriel GmbH &Co. KG, Darmstadt, Germany), whereas disturbingreflections were cut off. A telecentric lens (SILLOptics GmbH & Co. KG, Wendelstein, Germany)with a fixed magnification of 1.422 was used incombination with a CCD sensor (1280 ¥ 1024pixels, double-shutter, Sensicam PCO AG, Kelheim,Germany) to record the particle images in the lightsheet. Measurements were taken in the axial center-plane and in planes with axial offset.An infrared lightbarrier allows to synchronize the laser pulse to anydesired angular phase position of the impellerelectronically. A Labview controller console in com-bination with a counter card (National Instruments,Austin, TX, USA) was used for phase-lockedrecording. The pulse separation was 10 ms. Velocityfields were calculated with our own custom routinesby cross-correlation procedure with window shiftingand window refinement technique. The particledensity was high enough to achieve high-qualityvector fields with 32 ¥ 32 interrogation areas (7). Thefinal results represent averages of 50 DPIV record-ings at the same phase position. The velocity datawere used for direct comparison with the results fromthe numerical simulation.
Numerical setupThe numerical flow simulations of the pump allow
visualizing flow regions that are difficult to access byexperimental methods such as the radial gap part ofthe bypass (washout) channel. Furthermore, theyenable to show distributions, for example, of the wallshear stress or pressure in order to identify hot spots,and most important for the present work, the exactdetermination of integral quantities like the netwashout flow and the average pressures in predefinedsections of the flow field.
In the present study, the three-dimensional CADmodel of the blood pump system is transferred into acomputational structured multiblock grid using a gridgeneration tool (ANSYS ICEM CFD 10.0, ANSYS,Inc., Canonsburg, PA, USA). The computationaldomain as marked in Fig. 5 includes the extendedinflow region, the impeller, and the downstreamannular flow channel.
Special attention was paid to the meshing of thewashout holes. Refinements toward near-wall regionsand especially a fine discretization in blade and gapregions were taken into consideration. The grid con-sists in total of 2.65 million hexahedral cells. Thenumerical flow simulation was performed using thecommercial CFD software STAR-CD (Computa-tional Dynamics Limited, London, UK). The codenumerically solves the conservation equations ofmass and momentum by means of a finite volumeapproach. The Reynolds number based on theinterior diameter of the pump casing Dref = 6.8 mmreaches values dependent on the working point ofabout 2875. Therefore, turbulent flow is taken intoaccount by a Reynolds averaged approach and thelow-Reynolds k-e turbulence model. A major advan-tage of this model is its treatment of the near-wallregion. Blood was treated as a single-phase, incom-pressible, isothermal (37°C) Newtonian fluid withconstant density (1060 kg/m3) and viscosity (3.6 cP).Boundary conditions were chosen in agreement withthe experimental situation described in the sectionbefore and are defined as follows: a fully developedturbulent velocity profile is set at the entrance of theinflow cannula. Due to the incompressible characterof the fluid, the pressure was set in average constantin the outlet of the test compartment, so that thesimulated relative pressure field can be transferred inthe postprocessing to the correct pressure level with
FIG. 4. Optical arrangement (direction of view [DOV]) (4–6).
Impeller
FIG. 5. Domain of computation.
WASHOUT EFFECT IN A MAGNETICALLY DRIVEN AXIAL BLOOD PUMP 3
JOBNAME: No Job Name PAGE: 4 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
the help of experimental measurements. The simula-tions were carried out on a static grid with additionalsource terms due to rotation in the conservationequations, which turned out to be very adequate forthe purpose of this study. The stagnation point at thetip of the rotor’s hub is defined as the origin of thecoordinate system as shown in Fig. 6. Position P1 (seeFig. 5) is located 2.95 Dref in front of the origin, andposition P2 is located 3.5 Dref behind the origin. Thex–z plane is defined here as the plane perpendicularto the washout hole.
RESULTS
The results are structured as follows: first, the simu-lated operating points are compared against mea-sured performance curves and are discussed, then,the washout flow is determined and analyzed in moredetail for different impeller/motor gap widths, thecomputed general flow field is described and com-pared with experimental data, and finally, the liftoffforce and minimum static pressure are derived.
Flow curvesThe chosen working points for simulations were
extracted from hydraulically measured performancecurves obtained over the whole clinical pump system.The overall pressure differences were varied from 60to 120 mm Hg, the rotational speeds from 20 000 to28 000 rpm. The gap between the impeller and themotor was 400-m wide. The comparison of the ana-lyzed working points within the operating envelopeof the pump is shown in Fig. 7. Red and blue denotemeasured and computed values, respectively. Thepressure heads in the curves of Fig. 7 were obtainedexperimentally and numerically from the averagepressures acting in sections close to the impeller P1and P2 as introduced in Fig. 5 for direct comparison.The curves show good agreement.
A useful way to compare all cases is to nondimen-sionalize the pressure head DpP1-P2 between positionsP1 and P2 from Fig. 5 and the volume flow Q in thefollowing way:
Head coefficient: ψρ
= −Δp
u
P P
tip
2 1
212
Flow coefficient: ϕ π= Q
d u4
2tip tip
where12
2ρutip means a fictitious dynamic pressure,
andπ4
2d utip tip a fictitious volume flow based on the
blade tip speed, utip = pndtip, and the blade tip diam-eter, dtip. The diagram in Fig. 8 confirms the expectedoverlapping of the curves corresponding to differentrotational speeds n as in turbomachinery theory (8).
Washout flowThe washout flow acts as a bypass that permits
recirculation of the blood when a certain bypass pres-sure difference between both ends of the washoutchannel (dependent on the rotational speed of theimpeller) is reached. In this section, the axial gapbetween the impeller and the motor will be variedfrom 400 to 600 m as shown in Fig. 9.
FIG. 6. Side view of the region of interest.
0
50
100
150
200
250
300
0 1 2 3 4
flow [l/min]
Dp2-
1 [
mm
Hg
]
20000 m24000 m28000 m20000 c24000 c28000 c
FIG. 7. Comparison of flow curves: measured (m) and computed(c).
JOBNAME: No Job Name PAGE: 5 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
The upper limit of 600 m ensures sufficient mag-netic attraction, which prevents the impeller to liftoff. The net liftoff force is the sum of all fluiddynamicforces acting onto the impeller’s surface, for example,the pressure and viscous forces. In order to avoid adamaging physical contact between the impeller andthe motor at possible tumbling instabilities of therotor before reaching the actual working point, thelowest tolerable axial gap width is set to 400 m.
In general, the washout flow is a function of thechannel geometry as well as of the operation condi-tions such as the rotational speed and the pressurehead across the pump. Figure 10 shows the washoutflow results obtained for three different axial gapwidths at an overall pressure head of Dp = 80 mm Hg.A gap width of 500 m evidenced the highest washoutflow for all working conditions considered.
On the one hand, for a wider gap, the pressure lossbecomes higher due to secondary flows; on the otherhand, for a smaller gap width, the pressure loss gets
higher due to friction. In effect, a trade-off betweenthese two tendencies leads to the optimal gap widthof 500 m.
In Fig. 11, the whole flow field in a plane throughthe washout hole and in a perpendicular plane in theimpeller region is shown. The streamlines show theexistence of a stagnation region near the pivot (seealso Fig. 2).
The tip leakage vortex, a back-flow region at theimpeller housing, and the secondary flow in the axialgap between the impeller and the motor are alsovisualized. In addition, the plots include the relativepressure distributions, which are color coded accord-ing to the color map. The reference pressure for theCFD simulations was the atmospheric pressure of750 mm Hg, which was held constant at the outletboundary. The driving pressure difference for thewashout flow can be deduced from the plots.
400 m £ h £ 600 m
Magnetic Compensation
Lift-off
FIG. 9. Washout gap geometry variation and net force direction.
30
40
50
60
70
80
20000 24000 28000rotational speed [rpm]
wa
sh
ou
t fl
ow
[m
l/min
]
400 microns
500 microns
600 microns
FIG. 10. Washout flow at different rotational speeds and axialgap widths; overall pressure head Dp = 80 mm Hg.
FIG. 11. Sectional streamlines pattern andpressure distribution in the impeller regionfor working condition: n = 24 000 rpm,Q = 2.81 L/min, Dp = 80 mm Hg.
prel [mmHg]
Motor Impeller
Fig. 12 left
Flow Recirculation Secondary Flow
Tip Leakage Vortex
WASHOUT EFFECT IN A MAGNETICALLY DRIVEN AXIAL BLOOD PUMP 5
Artif Organs, Vol. ••, No. ••, 2008
1
2
3456789
10111213141516171819202122
2324
25
262728293031323334353637383940414243
44
45464748
m101
colour online & in print
m101
colour online & in print
m101
colour online & in print
JOBNAME: No Job Name PAGE: 6 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
Comparison CFD to DPIVFigure 12 shows a comparison between numerical
simulation and experiment of the flow field in thewashout hole region. The streamlines (left) and thevelocity vectors (right) evidence the path of the massleaving the radial gap through the washout hole. Thebypass flow is sharply turned in downstream direc-tion by the oncoming main flow, indicating that themajor part leaves the hole at the trailing edge.
Figure 13 highlights the inward direction of theflow in an axial section plane near the front end of themotor.The slope of the paths of the particle traces (3)agrees well with the slope of the streamlines.
Liftoff force and minimum static pressureFrom the numerical simulations, we were further
able to calculate the net liftoff force and to localizelow-pressure regions. For an overall pressure headof Dp = 80 mm Hg and a flow of 2.81 L/min at24 000 rpm, the liftoff force is 0.38 N. The MicroVadblood pump system is designed to operate withoutany risk of cavitation within the operationalenvelope. To obtain the absolute pressure field pabs,the computed absolute pressure level pcomp,abs isadapted according to the experimental pressurepexp,P1 and the computed pressure pcomp,P1 at positionP1 as follows: pabs = pcomp,abs + (pexp,P1 - pcomp,P1).
The results of the CFD study showed that thelowest static pressure is 333.8 mm Hg at the operat-
ing point of 28.000 rpm and 60 mm Hg. This highestdepression occurs at the suction side of the bladesnear the leading edge (Fig. 14). The volume- andarea-based distributions of the pressure in the impel-ler region for this specific case are given in Fig. 15.They indicate that cavitation will not occur, whenassuming a vapor pressure of 47 mm Hg for blood asgiven by Chambers et al. (9). The present assistdevice is an intracorporeal blood pump, so that nucleisize distributions and concentrations in the blood arenot likely to differ from those encountered in vivo.
FIG. 13. Flow and pressure distributionin the axial gap between impeller andmotor; comparison of numerical simulation(left) and experiment (right): streamlines(left, Dp = 120 mm Hg, n = 24 000 rpm,Q = 2.1 L/min) versus particle traces(right, Dp = 135 mm Hg, n = 24 000 rpm,Q = 1.8 L/min).
prel [mmHg]
Pabs [mmHg]
FIG. 14. Pressure distribution on the impeller for n = 28.000 rpmand Dp = 60 mm Hg.
M. TRIEP ET AL.6
Artif Organs, Vol. ••, No. ••, 2008
123456789
1011121314151617181920212223242526272829
3031323334
35
3637383940414243
44
4546474849505152535455
56
5758
m101
colour online & in print
m101
colour online & in print
JOBNAME: No Job Name PAGE: 7 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
DISCUSSION AND CONCLUSION
The main goal of the present investigation was thestudy and optimization of the induced washout flowthrough the axial impeller–motor gap and the radialimpeller–axle gap in order to avoid thrombus forma-tion during all operation conditions of the seallessCircuLite pump. Steady-state, three-dimensionalturbulent simulations and DPIV measurementswere carried out for various working conditions. Theresults obtained via the CFD simulations are in agood qualitative and quantitative agreement with theDPIV measurements.The washout flow is maximizedwith the axial impeller–motor gap at 500 m. Anincrease of the free cross-section of the gap leads to adecrease of the pressure loss and therefore to anincrease of the washout flow. The washout channelacts as a classical bypass in pumping systems. There-fore, any increase in the bypass flow (washout) coun-teracts to the pressure load and net flow rate at theworking point.
Furthermore, the liftoff force obtained via CFD iswell below the value for the magnetic attractiongiven by Kerkhoffs et al. (3), and the minimum staticpressure emerging in the pump is well above thevapor pressure of blood of 47 mm Hg.
In the next step, continued iterative geometricadaptations for an even better washout effect atphysiological pressure loads will be considered. It isimportant to include the study of fluid residence timeand the effectiveness of the mass exchange in thefront part of the pivot.
In addition, the mechanical stability of the pumpwill be studied, in particular, the restoring momentfrom the pressure acting on the impeller’s surface. Ineffect, a trade-off of the mechanical stability, on theone hand, and a good washout effect, on the otherhand, has to be established.
REFERENCES
1. Havlik R, Kerkhoffs W, Jiao L, Schumacher O, Reul H, HabibN. Intravascular micropump for augmented liver perfusion: firstin vivo experience. Artif Organs 2001;25:392–4.
2. Mahmood A, Kerkhoffs W, Schumacher O, Reul H. Investiga-tion of materials for blood-immersed bearings in a microaxialblood pump. Artif Organs 2003;27:169–73.
3. Kerkhoffs W, Schumacher O, Meyns B, et al. Design, develop-ment, and first in vivo results of an implantable ventricular assistdevice, MicroVad. Artif Organs 2004;28:904–10.
4. Brücker C, Schröder W, Apel J, Reul H, Siess T. DPIV study ofthe flow in a microaxial blood pump. Proceedings of the 9thSymposium on Transport Phenomena and Dynamics of Rotat-ing Machinery. ISROMAC-9. Honolulu, HI, 2002.
5. Triep M, Brücker C, Schröder W, Siess T. Computational fluiddynamics and digital particle image velocimetry study of theflow through an optimized micro-axial blood pump. ArtifOrgans 2006:30:384–91.
6. Triep M, Brücker C, Siess T. DPIV-measurements of the flowfield in a micro-axial blood pump. Proceedings of the 13th Sym-posium on Applications of Laser Techniques to Fluid Mechanics.Lisbon, 2006.
7. Adrian R. Particle-imaging techniques for experimental fluidmechanics. Ann Rev Fluid Mech 1991;23:261–301.
8. Gülich J. Kreiselpumpen: Handbuch für Entwicklung, Anlagen-planung und Betrieb, 2nd Edition. Berlin, Germany: Springer-Verlag, 2004.
9. Chambers S, Bartlett R, Ceccio S. Determination of the in vivocavitation nuclei characteristics of blood. ASAIO J 1999;45:541–9.
FIG. 15. Volume- (left) and area (right)-based distributions of pressure for n =28.000 rpm and Dp = 60 mm Hg.
Pressure distribution
WASHOUT EFFECT IN A MAGNETICALLY DRIVEN AXIAL BLOOD PUMP 7
JOBNAME: No Job Name PAGE: 8 SESS: 20 OUTPUT: Tue Sep 2 16:41:29 2008/v2503/blackwell/journals/aor_v32_i10/aor_630
SNP Best-set Typesetter Ltd.Journal Code: AOR Proofreader: EmilyArticle No: 630 Delivery date: 2 September 2008Page Extent: 7 Copyeditor: Gem
AUTHOR QUERY FORM
Dear Author,During the preparation of your manuscript for publication, the questions listed below have arisen. Please
attend to these matters and return this form with your proof.Many thanks for your assistance.
QueryReferences
Query Remark
1 Au: Please note that all instances of “digital particle-image velocimetry” havebeen changed to “digital particle image velocimetry” for consistencythroughout the text. Please confirm if this is correct.
2 Au: The location details “Saddle Brook, NJ, USA” has been added forCircuLite. Please confirm if this is correct.
3 Au: Please supply the manufacturer details of MicroVad pump system.
4 Au: Please clarify if all instances of “28.000 rpm” should be changed to“28 000 rpm”.
5 Au: Should CAD be written in full? If so, please provide the full form.
6 Au: Please note that the city “Canonsburg” has been added for ANSYS, Inc.Please confirm if this is correct.
